This paper presents a novel and computationally efficient deep learning-based object detection framework that fuses RGB images and LiDAR point cloud data to address persistent challenges in autonomous driving scenarios such as occlusion, varying lighting conditions, and complex object geometries. The proposed model, termed SimpleFusionModel, introduces an intermediate-level feature fusion strategy, where modality-specific features extracted via ResNet-50 (for RGB images) and PointNet++ (for LiDAR point clouds) are combined to form a unified representation before bounding box prediction. This fusion approach balances semantic richness and spatial precision while maintaining real-time feasibility. Unlike prior works that rely on early or late fusion, or transformer-heavy architectures like TransFusion, our method provides a lightweight yet accurate alternative. The system is trained and evaluated on the KITTI 3D object detection dataset, and its performance is benchmarked against several state-of-the-art methods including YOLOv3, PointPillars, MV3D, and TransFusion. Evaluation using metrics such as Intersection over Union (IoU), Precision, Recall, Average Precision (AP), and F1-score demonstrates that SimpleFusionModel achieves superior detection accuracy with low variance across multiple runs, particularly in detecting challenging classes like pedestrians and cyclists. Additionally, the model maintains competitive inference speed (27 FPS), highlighting its suitability for real-time deployment in autonomous vehicles.